osu! Beatmap Stats

Introduction

Basic data visualization of beatmap info provided by osu! API. https://github.com/ppy/osu-api/wiki

Only uses ranked/loved/qualified maps. Graphs focusing on standard mode.

(A few maps, due to ranked/loved irregularities, are actually in graveyard but included in the API query anyway. Ex. https://osu.ppy.sh/b/766190&m=2 with 1 loved CtB diff. )

library(ggplot2)
library(gridExtra)
library(plyr)
library(jsonlite)
library(varhandle)
library(chron)
library(magrittr)

# Cached chunk
# Import data from JSON and remove duplicate rows
beatmaps <- unique(do.call("rbind", fromJSON("maps.json")))

# Set global figure width and height
knitr::opts_chunk$set(fig.width=10, fig.height=6) 
# Disable warnings and messages
knitr::opts_chunk$set(message = FALSE, warning = FALSE)

# Convert strings of ints and floats to numeric datatypes
beatmaps.isintcol <- sapply(beatmaps, function(col) all(check.numeric(col, only.integer=TRUE))) 
beatmaps.isnumcol <- sapply(beatmaps, function(col) all(check.numeric(col))) & !beatmaps.isintcol
beatmaps[, beatmaps.isintcol] = sapply(beatmaps[, beatmaps.isintcol], as.integer)
beatmaps[, beatmaps.isnumcol] = sapply(beatmaps[, beatmaps.isnumcol], as.numeric)

# Convert MySQL datetimes to R datetimes. As far as the new site goes, this appears to be UTC-4, but
# this may be a local thing. Probably broken. 
beatmaps$approved_date <- as.POSIXct(beatmaps$approved_date, tz="Etc/GMT+4")
beatmaps$last_update <- as.POSIXct(beatmaps$last_update, tz="Etc/GMT+4")


# Create labels and data frames for each gamemode
gamemodes <- c("std", "taiko", "ctb", "mania")
gamemode.labels <- c("Standard", "Taiko", "CtB", "Mania")
beatmaps$mode <- factor(beatmaps$mode, labels=gamemode.labels)
for (i in 1:4) {
  assign(gamemodes[i], beatmaps[beatmaps$mode == gamemode.labels[i],])
}

# Various plot parameters for convenience
# These will usually leave a few outlier maps out
AR_y_scale          <- scale_y_continuous(breaks=seq(0,10))
SR_y_scale          <- scale_y_continuous(limits=c(0,10), breaks=seq(0,10,1))
approved_x_scale    <- scale_x_datetime(date_breaks="1 year", date_labels="%Y")

legend_title_fill   <- labs(fill="Mode")
legend_title_color  <- labs(color="Mode")

# Center titles
theme_update(plot.title = element_text(hjust = 0.5))

Plots

Star Rating

diff_x_scale <- scale_x_continuous(limits=c(0,10), breaks=seq(0,10,0.5))
diff_hist <- geom_histogram(binwidth=0.05)

# Histogram of star rating (all modes)
ggplot(beatmaps, aes(difficultyrating, fill=as.factor(mode))) + 
  ggtitle("Total Star Rating (All Modes)") + legend_title_fill +
  diff_x_scale + diff_hist

# Frequency polygon of SR (all modes)
ggplot(beatmaps, aes(difficultyrating, color=as.factor(mode))) + 
  ggtitle("Star Rating (All Modes)") + legend_title_color +
  diff_x_scale + geom_freqpoly(binwidth=0.05) 

# Histograms of SR (all modes) with separate y scales
diffplots <- llply(1:4, function(m) {
  ggplot(beatmaps[as.numeric(beatmaps$mode) == m, ], aes(difficultyrating)) +
    diff_x_scale + diff_hist + 
    ggtitle(gamemode.labels[m])
})
grid.arrange(grobs = diffplots, top="Star Rating Distributions (All Modes)")

The ranking criteria until recently required non-marathon maps to have a difficulty Normal or below. According to the new criteria, any map with less than 3:30 drain time requires a Normal or below. This explains the large quantity of maps between 1* and 2*.

After these Easy and Normal maps, the most popular standard maps are between 3* and 4*. There is a small but noticeable spike in standard maps at about 5.25*.

Taiko stands out for having the most unimodal-looking distribution.

Total Length

length_x_scale <- scale_x_continuous(limits=c(0,600), breaks=seq(0,600,30))
length_hist <- geom_histogram(binwidth=1)
x_labels_90 <- theme(axis.text.x = element_text(angle=90, hjust=1))

# Histogram of total length (all modes)
ggplot(beatmaps, aes(total_length, fill=as.factor(mode))) + 
  ggtitle("Total Beatmap Length (All Modes)") + legend_title_fill +
  length_x_scale + length_hist

# Frequency polygon of total length (all modes)
ggplot(beatmaps, aes(total_length, color=as.factor(mode))) + 
  ggtitle("Beatmap Length (All Modes)") + legend_title_color +
  length_x_scale + geom_freqpoly(binwidth=1)

# Histograms of total length (all modes)
lengthplots <- llply(1:4, function(m) {
  ggplot(beatmaps[as.numeric(beatmaps$mode) == m, ], aes(total_length)) +
    length_hist + length_x_scale + x_labels_90 + 
    ggtitle(gamemode.labels[m])
})

grid.arrange(grobs=lengthplots, top="Beatmap Length Distributions (All Modes)")

The massive spike in maps 85-90 seconds long corresponds to the ever-popular TV Size map genre.

This occurs in every gamemode, though mania and taiko have relatively large numbers of two minute maps.

Playcount

# Frequency polygons of playcount (all modes)
ggplot(beatmaps, aes(playcount, color=as.factor(mode))) + 
  ggtitle("Playcount (All Modes)") + legend_title_color +
  scale_x_continuous(limits=c(0,1000000)) + 
  geom_freqpoly(binwidth=5000)

Date approved

beatmaps$month <- cut(beatmaps$approved_date, breaks="month")
year_x_scale <- scale_x_discrete(breaks=unique(cut(beatmaps$approved_date, breaks="year"))) 
x_labels_45 <- theme(axis.text.x = element_text(angle=45, hjust=1))

# Bar chart of date approved (all modes)
ggplot(beatmaps, aes(x=month, fill=mode)) + 
  ggtitle("Date Approved (All Modes)") +
  geom_bar(width=1) + 
  year_x_scale + x_labels_45

# Frequency polygon of date approved (all modes)
ggplot(beatmaps, aes(x=month, group=mode, color=mode)) +
  ggtitle("Date Approved (All Modes)") + 
  legend_title_color +
  geom_freqpoly(stat="count") +
  year_x_scale + x_labels_45

Tables

Most frequent artists, titles, sources, and creators

library(knitr)
most.frequent.kable <- function(arr, lab) {
  # Neat use of pipe
  arr %>% table %>% sort(decreasing=TRUE) %>% head(20) %>% kable(col.names=c(lab, "Freq"))
}

beatmaps$artist  %>% most.frequent.kable("Artist")

Artist	Freq
Hatsune Miku	653
ClariS	421
KOTOKO	398
fripSide	362
xi	349
senya	339
IOSYS	331
Various Artists	329
LiSA	323
yanaginagi	311
Camellia	309
ZUN	272
Duca	257
Rita	255
M2U	245
Chata	226
u’s	225
Hanatan	224
nano	224
Suzuki Konomi	211

beatmaps$title   %>% most.frequent.kable("Title")

Title	Freq
Piano 7K BMS Pack	193
Piano Beatmap Set	114
Harumachi Clover	70
Granat	65
Ai no Scenario	62
PEPPY FIX TAIKO STAR RATING PLEASE for a happier	61
Tokyo (Innovaderz Remix)	61
MIIRO	58
Hitorigoto -TV MIX-	55
Re:TrymenT	54
Haru Modoki	52
Natsukoi Hanabi	52
TSLove	52
Contrail Kiseki	51
Paradisus-Paradoxum	51
Untan Goose	51
Uso no Hibana	50
Six Trillion Years and Overnight Story	49
Gabriel Drop Kick	43
Oriental Blossom	42

beatmaps$source  %>% most.frequent.kable("Source")

Source	Freq
	24652
Touhou	2370
BMS	1557
東方Project	754
SOUND VOLTEX III GRAVITY WARS	466
DJMAX	455
beatmania IIDX	398
SOUND VOLTEX II -infinite infection-	390
osu!	258
Vocaloid	242
Taiko no Tatsujin	222
Deemo	202
jubeat	187
Nico Nico Douga	178
Love Live! School idol project	174
SOUND VOLTEX BOOTH	169
REFLEC BEAT groovin’!!	166
Cytus	164
K-ON!!	153
艦隊これくしょん -艦これ-	146

beatmaps$creator %>% most.frequent.kable("Creator")

Creator	Freq
osuplayer111	572
Sotarks	571
DJPop	563
tutuhaha	387
ztrot	377
Larto	354
Natsu	345
Monstrata	331
Ascendance	311
pishifat	311
Lasse	310
Gero	304
wcx19911123	293
Milan-	279
ouranhshc	262
alacat	248
NatsumeRin	247
MoonFragrance	234
James	232
Fycho	221

Most favorited mapsets

beatmaps %>% 
  subset(!duplicated(beatmaps$beatmapset_id)) %>%  # Keep rows with unique beatmapset_id
  arrange(desc(favourite_count)) %>% 
  head(50) %>% `[`(c("creator", "artist", "title", "favourite_count")) %>% kable

creator	artist	title	favourite_count
W h i t e	Kuba Oms	My Love	13163
Fort	Panda Eyes & Teminite	Highscore	10292
jonathanlfj	cYsmix feat. Emmy	Tear Rain	9763
Charles445	Rostik	Liquid (Paul Rosenthal Remix)	8282
VINXIS	Reol	No title	7604
Ekoro	UNDEAD CORPORATION	Everything will freeze	6886
Kuria	Linked Horizon	Guren no Yumiya (TV Size)	6880
Doormat	ClariS	Hitorigoto -TV MIX-	6606
Awaken	Konuko	Toumei Elegy	6534
Voltaeyx	TheFatRat	Mayday (feat. Laura Brehm)	6335
Takuya	S3RL	Pika Girl	5510
Bearizm	Station Earth	Cold Green Eyes ft. Roos Denayer	5425
Saten-san	Yousei Teikoku	Kokou no Sousei	5300
gowww	Hatsune Miku & Megpoid Gumi	MATRYOSHKA	4878
h3k1ru	Yiruma & Skullee	River Flows In You (A Love Note)	4849
ouranhshc	Masayoshi Minoshima feat. nomico	Bad Apple!!	4829
-kevincela-	Rameses B	Flaklypa	4829
ktgster	Chasers	Lost	4770
Sekai-nyan	Suzuki Konomi	This game (TV Size)	4618
Kuria	ONE OK ROCK	Answer is Near	4591
eLy	Feint	Tower Of Heaven (You Are Slaves)	4461
Secretpipe	S3RL	Bass Slut (Original Mix)	4399
Monstrata	RADWIMPS	Zen Zen Zense (movie ver.)	4352
Natsu	Hanatan	Airman ga Taosenai (SOUND HOLIC Ver.)	4328
NatsumeRin	Hatsune Miku	Senbonzakura (Short Ver.)	3966
kristi71111	TK from Ling tosite sigure	unravel (TV edit)	3942
Rue	DJ Genericname	Dear You	3886
Monstrata	Porter Robinson & Madeon	Shelter	3860
Asphyxia	xi	Blue Zenith	3849
Kagetsu	KANA-BOON	Silhouette	3839
osuplayer111	Getter Jaani	Rockefeller Street (Nightcore Mix)	3703
Smoothie	UNDEAD CORPORATION	Yoru Naku Usagi wa Yume o Miru	3677
Star Stream	Wotamin	Gigantic O.T.N	3627
Kyshiro	toby fox	MEGALOVANIA	3613
Blue Dragon	The Quick Brown Fox	The Big Black	3403
RLC	Himeringo	Yotsuya-san ni Yoroshiku	3377
Multiple Creators	Soleily	Renatus	3345
Sherry	Nanahira	Frightfully-insane Flan-chan’s frightful song	3327
jonathanlfj	Reol	Plus Danshi ver Reol	3324
Len	Shawn Wasabi	Marble Soda	3288
Ephemeral	Masayoshi Minoshima ft. nomico	Bad Apple!!	3219
Tarrasky	Agnete Kjolsrud	Get Jinxed	3165
Guy	Aoi Eir	IGNITE (TV size ver.)	3119
AllStar12	yuikonnu & ayaponzu*	Super Nuko World	3033
Garven	Saiya	Remote Control	3028
handsome	Reol	MONSTER	3009
Gaia	Reol	Asymmetry	2993
rui	Hatsune Miku	Rubik’s Cube	2967
Monstrata	9mm Parabellum Bullet	Inferno	2959
Jacob	NOMA	Brain Power	2908

Most played maps

beatmaps %>%
  arrange(desc(playcount)) %>% 
  head(50) %>% `[`(c("creator", "artist", "title", "version", "playcount")) %>% kable

creator	artist	title	version	playcount
W h i t e	Kuba Oms	My Love	Hard	21429513
W h i t e	Kuba Oms	My Love	Normal	20405157
jonathanlfj	cYsmix feat. Emmy	Tear Rain	Normal	18188427
Blue Dragon	The Quick Brown Fox	The Big Black	WHO’S AFRAID OF THE BIG BLACK	14209863
jonathanlfj	cYsmix feat. Emmy	Tear Rain	Hard	12840626
ktgster	Chasers	Lost	Normal	12456963
-kevincela-	Rameses B	Flaklypa	Normal	12124613
Charles445	Rostik	Liquid (Paul Rosenthal Remix)	Easy	10881187
Blue Dragon	Team Nekokan	Can’t Defeat Airman	Holy Shit! It’s Airman!!	10878775
-kevincela-	Rameses B	Flaklypa	Hard	10611576
W h i t e	Kuba Oms	My Love	Insane	9287944
Multiple Creators	Soleily	Renatus	Normal	8065171
jonathanlfj	cYsmix feat. Emmy	Tear Rain	Insane	8028482
ktgster	Chasers	Lost	Hard	7715600
Rue	DJ Genericname	Dear You	Dear Rue	7548246
Bearizm	Station Earth	Cold Green Eyes ft. Roos Denayer	Divine	7343941
h3k1ru	Yiruma & Skullee	River Flows In You (A Love Note)	Love Note	7328010
VINXIS	Reol	No title	Light Insane	6820213
Charles445	Rostik	Liquid (Paul Rosenthal Remix)	Normal	6818829
val0108	Lily	Scarlet Rose	0108 style	6684374
Charles445	Rostik	Liquid (Paul Rosenthal Remix)	Hard	6659517
Ekoro	UNDEAD CORPORATION	Everything will freeze	Insane	6604128
Reikin	Nico Nico Douga	U.N. Owen Was Her?	Normal	6495851
Fort	Panda Eyes & Teminite	Highscore	Another	6442871
Fort	Panda Eyes & Teminite	Highscore	LGV’s Insane	6255646
Lust	Tsunamaru	Daidai Genome	Insane	6182514
Takuya	S3RL	Pika Girl	Hard	5910259
Saten-san	Yousei Teikoku	Kokou no Sousei	Hard	5760621
Nakagawa-Kanon	xi	FREEDOM DiVE	Another	5520280
Kuria	Linked Horizon	Guren no Yumiya (TV Size)	DS’s Hard	5517593
Flask	Fujijo Seitokai Shikkou-bu	Best FriendS -TV Size-	Fycho’s Insane	5498716
VINXIS	Reol	No title	byfaR’s Hard	5419243
Garven	Saiya	Remote Control	Insane	5395495
xxdeathx	FLOWxGRANRODEO	7 -seven- -TV SIZE -	Expert	5357990
Damnae	raja	the light	Normal	5289914
Kuria	Linked Horizon	Guren no Yumiya (TV Size)	alacat’s Normal	5244344
Fort	Panda Eyes & Teminite	Highscore	Hyper	5227173
Doormat	ClariS	Hitorigoto -TV MIX-	Insane	5173748
Bearizm	Station Earth	Cold Green Eyes ft. Roos Denayer	apple’s Insane	5172218
galvenize	DJ Fresh	Gold Dust	Insane	5150833
Taeyang	kradness&Reol	Remote Control	Max Control!	5141365
Luerxa	Primastella	Koigokoro	Delis’ Insane	5138932
JauiPlaY	DJ Okawari	Flower Dance	Flower	5138709
Multiple Creators	Soleily	Renatus	Hard	5085398
val0108	Hatsune Miku	Mythologia’s End	Myth0108ia	5076064
Takuya	S3RL	Pika Girl	Insane	5024386
Natsu	Hanatan	Airman ga Taosenai (SOUND HOLIC Ver.)	Insane	4916733
Star Stream	Sagara Kokoro	Hoshizora no Ima	S.S	4879941
Nakagawa-Kanon	xi	FREEDOM DiVE	FOUR DIMENSIONS	4753427
Kuria	Linked Horizon	Guren no Yumiya (TV Size)	Insane	4679371

Scatterplots

# Scatterplot of AR vs BPM
ggplot(std, aes(bpm, diff_approach)) + 
  ggtitle("Approach Rate vs BPM") + 
  scale_x_continuous(limits=c(0,500)) + 
  AR_y_scale + 
  geom_point(alpha=0.1)

# Scatterplot of SR vs total length time
ggplot(std, aes(total_length, difficultyrating)) + 
  ggtitle("Star Rating vs Total Length") +
  length_x_scale + 
  SR_y_scale +
  geom_point(alpha=0.1)

# Scatterplot of max combo vs drain time
ggplot(std, aes(hit_length, max_combo)) + 
  ggtitle("Max Combo vs Drain Time") + 
  length_x_scale + 
  scale_y_continuous(limits=c(0,4000)) +
  geom_point(alpha=0.05)

# High linear correlation, as expected
summary(lm(max_combo ~ hit_length, data=std))

## 
## Call:
## lm(formula = max_combo ~ hit_length, data = std)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -3860.7  -134.3   -17.4   120.5 23020.5 
## 
## Coefficients:
##              Estimate Std. Error t value Pr(>|t|)    
## (Intercept) -99.20894    2.51272  -39.48   <2e-16 ***
## hit_length    4.89613    0.01715  285.53   <2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 272.8 on 56923 degrees of freedom
##   (4 observations deleted due to missingness)
## Multiple R-squared:  0.5888, Adjusted R-squared:  0.5888 
## F-statistic: 8.152e+04 on 1 and 56923 DF,  p-value: < 2.2e-16

# Scatterplot of favorite count vs playcount
ggplot(std, aes(playcount, favourite_count)) + 
  ggtitle("Favorite Count vs Playcount") + 
  scale_x_continuous(limits=c(0,1000000)) +
  scale_y_continuous(limits=c(0,1000)) +
  geom_point(alpha=0.05)

# Scatterplot of playcount vs total length
ggplot(std, aes(total_length, playcount)) +
  ggtitle("Playcount vs Total Length") + 
  length_x_scale + 
  scale_y_continuous(limits=c(0,1000000)) +
  geom_point(alpha=0.1)

# Scatterplot of AR vs date approved
ggplot(std, aes(approved_date, diff_approach)) + 
  ggtitle("Approach Rate vs Date Approved") + 
  AR_y_scale + 
  approved_x_scale + 
  geom_point(alpha=0.05)

# Scatterplot of SR vs date approved
ggplot(std, aes(approved_date, difficultyrating)) +
  ggtitle("Star Rating vs Date Approved") +
  SR_y_scale + 
  approved_x_scale + 
  geom_point(alpha=0.1)

Spread info

# Playcount by song time, categorized by spread icon
# https://osu.ppy.sh/help/wiki/Difficulties#star-rating Not sure about values between boundaries

spread.sr = c(0, 1.51, 2.26, 3.76, 5.26, 6.76)  
spread.names = c("Easy", "Normal", "Hard", "Insane", "Expert", "Expert+")
spread.colors = c("olivedrab3", "paleturquoise", "gold", "hotpink", "purple", "darkgray")

# Assign difficulty rating by spread ranges to spread names 
beatmaps$spread_name <- spread.names[cut(beatmaps$difficultyrating, spread.sr, right=FALSE, labels=FALSE)]
std <- beatmaps[beatmaps$mode == "Standard",]  # update std 

hitlength.bins = seq(0, 360, 30)
par(mfrow=c(2,3), mar=c(4,4,4,1), cex.main=2)
for (i in 1:length(spread.names)) {
  std.spread = std[std$spread_name == spread.names[i], ]
  playcount.bin.sum = sapply(split(std.spread, cut(std.spread$hit_length, hitlength.bins)),
                             function(df) sum(df$playcount))

  barplot(playcount.bin.sum, space=0, width=30, xlab="Hit length (s)", ylab="Playcount Total", main=spread.names[i],
          col=spread.colors[i], axisnames=FALSE)
  axis(1, at=hitlength.bins)
}

# Same but 150+ hitlength and stacked bars
hitlength.bins.150 = seq(150, 360, 30)
playcount.bin.mat = matrix(ncol=length(hitlength.bins.150)-1, nrow=length(spread.names))
colnames(playcount.bin.mat) = head(hitlength.bins.150, -1)
rownames(playcount.bin.mat) = spread.names

for (i in 1:nrow(playcount.bin.mat)) {
  std.spread = std[std$spread_name == spread.names[i], ]
  playcount.bin.mat[i,] = sapply(split(std.spread, cut(std.spread$hit_length, hitlength.bins.150)),
                                 function(df) sum(df$playcount))
}

dev.off()  # Reset par 

## null device 
##           1

barplot(playcount.bin.mat, space=0, width=30, col=spread.colors, xlab="Hitlength (s)", ylab="Total Playcount",
        legend.text=spread.names, axisnames=FALSE, main="Total Playcount by Hitlength and Difficulty")
axis(1, at=hitlength.bins.150-hitlength.bins.150[1], labels=hitlength.bins.150)